On September 29, OpenAI introduced dots, AI agents that get their own cloud computer and can keep working after you leave the chat. They can connect to apps and follow an ongoing project.
The new part is not simply a longer answer. A dot might monitor feedback, update a document and ask for a decision when a person is needed. OpenAI says background research in connected apps is read-only. Sensitive actions should require approval, and users can set their own rules.
Access is still limited. Dots are rolling out to paying Pro and Business Premium customers in eligible markets. Enterprise customers can try a beta if their administrator enables it. OpenAI also says work with real consequences still needs review. A persistent helper may be useful, but a mistake can continue for longer too.
Some live demonstrations ran into problems during the launch. It is a useful reminder that an exciting new way of working begins as technology that still has to prove how dependable it is in everyday use.